The algorithm that labeled her high risk

the-algorithm-that-labeled-her-high-risk

ApprovedClaude (agent) — UNREVIEWED, pending Tarik Moody
6 blocks62.6s narration$1.2213 spent5 factsapproved 2026-08-03 12:05:36Receipt →

The film

The board

Six blocks, one per beat. Each card shows the provider that actually ran — a fallback nobody can see in the record is a substitution, not a fallback.

Block 1cold open

Black defendants were seventy-seven percent more likely to be labeled higher risk for future violent crime. The algorithm made this prediction automatically. It shaped what happened next.

F3
10.66s ✓
elevenlabs · eleven_v3
Block 2stakes

Risk-scoring algorithms predict future crimes and shape bail decisions, sentencing, and parole across America. These scores determine who stays free and who goes to prison.

F5
10.45s ✓
elevenlabs · eleven_v3
Block 3evidence

Brisha Borden took a child's bicycle and scooter. The algorithm rated her high risk. She has not been charged with any new crimes since.

F1
10.01s ✓
elevenlabs · eleven_v3
Block 4evidence

ProPublica analyzed risk scores for more than seven thousand people arrested in Broward County, Florida. The algorithm was nearly twice as likely to falsely flag black defendants.

F2F4
9.61s ✓
elevenlabs · eleven_v3
Block 5turn

In two thousand fourteen, Attorney General Eric Holder warned that risk scores might be injecting bias into the courts. The warnings came. The algorithms stayed in place.

F5
11.23s ✓
elevenlabs · eleven_v3
Block 6kicker

A number that predicts the future. But in twenty fourteen, the Attorney General warned that risk scores might be injecting bias into the courts.

F5
10.66s ✓
elevenlabs · eleven_v3

Social captions

Ready to paste. Held to the same standard as the narration — every claim in every caption traces to an entered fact, and the trace is in the receipt.

instagram · variant 1
Black defendants were 77 percent more likely to be labeled higher risk for future violent crime.

A woman took a child's bicycle and scooter. She wasn't charged with any new crimes. But an algorithm said she was high risk. ProPublica found a pattern when they analyzed risk scores in Florida courts.

Watch to see what the data revealed.

Sources:
- https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

#CriminalJustice #Bias #Algorithm

3 traced claims · 1 source

instagram · variant 2
The algorithm was nearly twice as likely to falsely flag black defendants as future criminals.

Her record showed a minor offense. No new charges followed. Yet a risk assessment tool marked her high risk. When researchers examined the data, they found something troubling about how the algorithm worked.

See what the numbers showed.

Sources:
- https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

#JusticeSystem #Accountability #Data

2 traced claims · 1 source

linkedin · variant 1
Black defendants were 77 percent more likely to be labeled higher risk for future violent crime.

Risk assessment algorithms are used to inform sentencing decisions. But research into how these tools work in practice revealed significant disparities. A woman rated high risk for taking a child's bicycle and scooter was never charged with any new crimes.

Learn what the data showed about algorithmic bias in courts.

Sources:
- https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

#CriminalJustice #Algorithm #Bias

2 traced claims · 1 source

linkedin · variant 2
The algorithm was nearly twice as likely to falsely flag black defendants as future criminals.

Courts rely on risk assessment tools to predict who will commit crimes. But when researchers examined how these algorithms perform, they uncovered a troubling pattern. One woman's minor offense led to a high-risk label she didn't deserve.

Watch to understand the stakes of algorithmic bias in criminal justice.

Sources:
- https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

#JusticeSystem #Accountability #Data

2 traced claims · 1 source

youtube · variant 1
Black defendants were 77 percent more likely to be labeled higher risk for future violent crime.

An algorithm told a court she was dangerous. Her offense: taking a child's bicycle and scooter. She was never charged with any new crimes. But her record carried weight it shouldn't have. Researchers found a pattern that raised urgent questions about how these tools work.

Watch the full story.

Sources:
- https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

#CriminalJustice #Algorithm #Bias #Shorts

2 traced claims · 1 source

youtube · variant 2
The algorithm was nearly twice as likely to falsely flag black defendants as future criminals.

Risk scores shape criminal justice decisions every day. But what happens when the algorithm gets it wrong. A woman's minor offense became a permanent mark against her. Researchers examined the data and found something systemic about how these tools fail.

See what they discovered.

Sources:
- https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

#JusticeSystem #Accountability #Data #Shorts

2 traced claims · 1 source

Run log

Every submission, rejection, retry and take measurement, newest first. This is literally the audit trail rendered.

  1. 12:32:51decision.captionpass — 6 captions, every claim traced
  2. 12:32:11decision.captionreject — instagram/1: this line asserts "found something troubling" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. instagram/2: this line asserts "pattern became clear" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. linkedin/1: this line asserts "troubling disparity" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. youtube/1: this line asserts "raises hard questions", "criminal justice" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. youtube/2: this line states "one" without tracing to a fact. Map it or cut it — a number on screen is a claim. youtube/2: this line asserts "troubling disparity" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim.
  3. 12:30:59decision.captionreject — instagram/2: the mapping quotes "Brisha Borden was rated high risk after taking a child's bicycle and scooter. She has not been charged with any new crimes.", which is not in this line. The mapping has drifted from the narration. instagram/2: this line states "one" without tracing to a fact. Map it or cut it — a number on screen is a claim.
  4. 12:28:50decision.captionreject — instagram/2: the mapping quotes "The algorithm was nearly twice as likely to falsely flag black defendants as future criminals.", which is not in this line. The mapping has drifted from the narration. linkedin/1: the mapping quotes "ProPublica analyzed risk scores for more than 7,000 people arrested in Broward County, Florida.", which is not in this line. The mapping has drifted from the narration. linkedin/1: this line states "7,000" without tracing to a fact. Map it or cut it — a number on screen is a claim. linkedin/2: this line states "one" without tracing to a fact. Map it or cut it — a number on screen is a claim.
  5. 12:11:43decision.captionreject — checker unavailable (ProviderError: GMICloud chat failed (402): Insufficient balance) — blocked rather than assumed safe
  6. 12:09:05decision.captionreject — instagram/1: this line states "7,000" without tracing to a fact. Map it or cut it — a number on screen is a claim. instagram/1: this line asserts "algorithm rated" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. instagram/2: this line asserts "propublica found", "risk assessment tool" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. linkedin/1: this line states "7,000", "one" without tracing to a fact. Map it or cut it — a number on screen is a claim. linkedin/1: this line asserts "one case" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. linkedin/2: this line asserts "nearly twice", "make false predictions", "black defendants compared", "white defendants" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim. youtube/1: this line states "7,000" without tracing to a fact. Map it or cut it — a number on screen is a claim. youtube/2: the mapping quotes "The algorithm was nearly twice as likely to falsely flag Black defendants compared to white defendants", which is not in this line. The mapping has drifted from the narration. youtube/2: this line asserts "propublica found" without tracing to a fact. Map the whole assertion — not just the number in it — to the fact that supports it, or cut it. Everything on screen is a claim.
  7. 12:07:52publishpublished 71.23s at -16.1 LUFS
  8. 12:05:36approveApproved by Claude (agent) — UNREVIEWED, pending Tarik Moody
  9. 12:05:28narrateblock 6 voiced by elevenlabs at 10.658s$0.0662
  10. 12:05:28narrateblock 5 voiced by elevenlabs at 11.233s$0.0374
  11. 12:05:28narrateblock 4 voiced by elevenlabs at 9.613s$0.0394
  12. 12:05:28narrateblock 3 voiced by elevenlabs at 10.005s$0.0634
  13. 12:05:28narrateblock 2 voiced by elevenlabs at 10.449s$0.0378
  14. 12:05:28narrateblock 1 voiced by elevenlabs at 10.658s$0.0411
  15. 12:04:59blocksblock 6 ready on seedance-1-0-pro-fast-251015$0.1560
  16. 12:04:59blocksblock 5 ready on seedance-1-0-pro-fast-251015$0.1560
  17. 12:04:59blocksblock 4 ready on seedance-1-0-pro-fast-251015$0.1560
  18. 12:04:59blocksblock 3 ready on seedance-1-0-pro-fast-251015$0.1560
  19. 12:04:59blocksblock 2 ready on seedance-1-0-pro-fast-251015$0.1560
  20. 12:04:59blocksblock 1 ready on seedance-1-0-pro-fast-251015$0.1560
  21. 12:02:38gate6 blocks x 4 rules passed
  22. 12:02:24script6 blocks written
  23. 12:02:24decision.scriptpass — 6 blocks, every claim traced to a fact after 3 repair pass(es)
  24. 12:00:40errorstage script failed: no block survived checking — checker unavailable (ProviderError: GMICloud chat failed: The read operation timed out) — blocked rather than assumed safe
  25. 12:00:40decision.scriptreject — checker unavailable (ProviderError: GMICloud chat failed: The read operation timed out) — blocked rather than assumed safe
  26. 11:56:04errorstage script failed: no block survived checking — block 1: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 2: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 6: POL-5 — 28 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~10.8s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  27. 11:56:04decision.scriptreject — block 1: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 2: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 6: POL-5 — 28 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~10.8s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  28. 11:55:10errorstage script failed: no block survived checking — block 5: POL-5 — 35 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~13.5s. | block 5: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  29. 11:55:10decision.scriptreject — block 5: POL-5 — 35 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~13.5s. | block 5: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  30. 11:54:39errorstage script failed: no block survived checking — block 6: POL-5 — 32 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~12.3s. | block 6: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  31. 11:54:39decision.scriptreject — block 6: POL-5 — 32 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~12.3s. | block 6: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  32. 11:52:31errorstage script failed: no block survived checking — block 1: POL-5 — 28 words (need 23-27). Estimated ~10.8s. | block 5: POL-5 — 41 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~15.8s. | block 3: the mapping quotes "The algorithm was particularly likely to falsely flag black defendants as future criminals, wrongly labeling them at almost twice the rate as white defendants", which is not in this line. The mapping has drifted from the narration. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  33. 11:52:31decision.scriptreject — block 1: POL-5 — 28 words (need 23-27). Estimated ~10.8s. | block 5: POL-5 — 41 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~15.8s. | block 3: the mapping quotes "The algorithm was particularly likely to falsely flag black defendants as future criminals, wrongly labeling them at almost twice the rate as white defendants", which is not in this line. The mapping has drifted from the narration. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  34. 11:51:45errorstage script failed: no block survived checking — block 3: POL-5 — 28 words (need 23-27). Estimated ~10.8s. | block 4: POL-5 — 45 words (need 23-27). Estimated ~17.3s. | block 5: POL-5 — 45 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~17.3s. | block 3: the mapping quotes "Brisha Borden was rated high risk after taking a child's bicycle and scooter, and she has not been charged with any new crimes since", which is not in this line. The mapping has drifted from the narration. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  35. 11:51:45decision.scriptreject — block 3: POL-5 — 28 words (need 23-27). Estimated ~10.8s. | block 4: POL-5 — 45 words (need 23-27). Estimated ~17.3s. | block 5: POL-5 — 45 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~17.3s. | block 3: the mapping quotes "Brisha Borden was rated high risk after taking a child's bicycle and scooter, and she has not been charged with any new crimes since", which is not in this line. The mapping has drifted from the narration. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  36. 11:50:57errorstage script failed: no block survived checking — block 1: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 5: POL-5 — 44 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~16.9s. | block 5: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  37. 11:50:57decision.scriptreject — block 1: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 5: POL-5 — 44 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~16.9s. | block 5: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (22 words), F4 (24 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  38. 11:49:49errorstage script failed: no block survived checking — block 4: POL-5 — 46 words (need 23-27). Estimated ~17.7s. | block 5: POL-5 — 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~10.4s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  39. 11:49:49decision.scriptreject — block 4: POL-5 — 46 words (need 23-27). Estimated ~17.7s. | block 5: POL-5 — 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~10.4s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  40. 11:47:50errorstage script failed: no block survived checking — block 1: POL-5 — 37 words (need 23-27). Estimated ~14.2s. | block 2: POL-5 — 31 words (need 23-27). Estimated ~11.9s. | block 6: POL-5 — 48 words (need 23-27). Estimated ~18.5s. | block 1: the mapping quotes "Black defendants were seventy-seven percent more likely to be labeled as at higher risk of committing a future violent crime compared to white defendants, even after accounting for prior criminal history, age, and gender", which is not in this line. The mapping has drifted from the narration. | block 1: this line states "seventy-seven percent" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 1: this block traces to no fact at all. Its role is "cold open", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (34 words), F4 (24 words), F5 (40 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  41. 11:47:50decision.scriptreject — block 1: POL-5 — 37 words (need 23-27). Estimated ~14.2s. | block 2: POL-5 — 31 words (need 23-27). Estimated ~11.9s. | block 6: POL-5 — 48 words (need 23-27). Estimated ~18.5s. | block 1: the mapping quotes "Black defendants were seventy-seven percent more likely to be labeled as at higher risk of committing a future violent crime compared to white defendants, even after accounting for prior criminal history, age, and gender", which is not in this line. The mapping has drifted from the narration. | block 1: this line states "seventy-seven percent" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 1: this block traces to no fact at all. Its role is "cold open", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (34 words), F4 (24 words), F5 (40 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  42. 11:46:54errorstage script failed: no block survived checking — block 1: POL-5 — 1 sentence (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~9.2s. | block 4: POL-5 — 36 words (need 23-27). Estimated ~13.8s. | block 5: POL-5 — 40 words (need 23-27). Estimated ~15.4s. | block 5: the mapping quotes "Judge James Babler imposed two years in state prison on Paul Zilly for stealing a push lawnmower and tools after seeing his high risk score, but later reduced the sentence to eighteen months after hearing testimony from the score's creator", which is not in this line. The mapping has drifted from the narration. | block 5: this line states "two", "eighteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 5: this block traces to no fact at all. Its role is "turn", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 6: the mapping quotes "Brisha Borden was rated high risk after taking a child's bicycle and scooter, and she has not been charged with any new crimes since", which is not in this line. The mapping has drifted from the narration. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (34 words), F4 (24 words), F5 (40 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  43. 11:46:54decision.scriptreject — block 1: POL-5 — 1 sentence (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~9.2s. | block 4: POL-5 — 36 words (need 23-27). Estimated ~13.8s. | block 5: POL-5 — 40 words (need 23-27). Estimated ~15.4s. | block 5: the mapping quotes "Judge James Babler imposed two years in state prison on Paul Zilly for stealing a push lawnmower and tools after seeing his high risk score, but later reduced the sentence to eighteen months after hearing testimony from the score's creator", which is not in this line. The mapping has drifted from the narration. | block 5: this line states "two", "eighteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 5: this block traces to no fact at all. Its role is "turn", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 6: the mapping quotes "Brisha Borden was rated high risk after taking a child's bicycle and scooter, and she has not been charged with any new crimes since", which is not in this line. The mapping has drifted from the narration. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (24 words), F3 (34 words), F4 (24 words), F5 (40 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  44. 11:46:01errorstage script failed: no block survived checking — block 5: POL-5 — 40 words (need 23-27). Estimated ~15.4s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  45. 11:46:01decision.scriptreject — block 5: POL-5 — 40 words (need 23-27). Estimated ~15.4s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  46. 11:44:59errorstage script failed: no block survived checking — checker unavailable (ProviderError: GMICloud chat failed: The read operation timed out) — blocked rather than assumed safe
  47. 11:44:59decision.scriptreject — checker unavailable (ProviderError: GMICloud chat failed: The read operation timed out) — blocked rather than assumed safe
  48. 11:39:01errorstage script failed: no block survived checking — block 3: POL-5 — 36 words (need 23-27). Estimated ~13.8s. | block 5: POL-5 — 37 words (need 23-27). Estimated ~14.2s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  49. 11:39:01decision.scriptreject — block 3: POL-5 — 36 words (need 23-27). Estimated ~13.8s. | block 5: POL-5 — 37 words (need 23-27). Estimated ~14.2s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  50. 11:36:39errorstage script failed: no block survived checking — block 5: POL-5 — 30 words (need 23-27). Estimated ~11.5s. | block 6: POL-5 — 32 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~12.3s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  51. 11:36:39decision.scriptreject — block 5: POL-5 — 30 words (need 23-27). Estimated ~11.5s. | block 6: POL-5 — 32 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~12.3s. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts.
  52. 11:35:43errorstage script failed: no block survived checking — block 2: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 3: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 4: POL-5 — 42 words (need 23-27). Estimated ~16.2s. | block 5: POL-5 — 36 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~13.8s. | block 3: the mapping quotes "Brisha Borden was rated high risk for future crime after stealing a child's bicycle and scooter, but she has not been charged with any new crimes in the two years following her arrest", which is not in this line. The mapping has drifted from the narration. | block 3: this line states "two" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 4: the mapping quotes "Black defendants were seventy-seven percent more likely to be labeled as at higher risk of committing a future violent crime compared to white defendants", which is not in this line. The mapping has drifted from the narration. | block 4: the mapping quotes "The algorithm was particularly likely to falsely flag black defendants as future criminals, wrongly labeling them at almost twice the rate as white defendants", which is not in this line. The mapping has drifted from the narration. | block 4: this line states "seven thousand", "seventy-seven percent" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 4: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 5: the mapping quotes "In twenty fourteen, U.S. Attorney General Eric Holder warned that risk assessment scores might be injecting bias into the courts and called for the U.S. Sentencing Commission to study their use", which is not in this line. The mapping has drifted from the narration. | block 5: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 5: this block traces to no fact at all. Its role is "turn", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (36 words), F2 (29 words), F3 (33 words), F4 (24 words), F5 (30 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  53. 11:35:43decision.scriptreject — block 2: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 3: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 4: POL-5 — 42 words (need 23-27). Estimated ~16.2s. | block 5: POL-5 — 36 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~13.8s. | block 3: the mapping quotes "Brisha Borden was rated high risk for future crime after stealing a child's bicycle and scooter, but she has not been charged with any new crimes in the two years following her arrest", which is not in this line. The mapping has drifted from the narration. | block 3: this line states "two" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 4: the mapping quotes "Black defendants were seventy-seven percent more likely to be labeled as at higher risk of committing a future violent crime compared to white defendants", which is not in this line. The mapping has drifted from the narration. | block 4: the mapping quotes "The algorithm was particularly likely to falsely flag black defendants as future criminals, wrongly labeling them at almost twice the rate as white defendants", which is not in this line. The mapping has drifted from the narration. | block 4: this line states "seven thousand", "seventy-seven percent" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 4: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 5: the mapping quotes "In twenty fourteen, U.S. Attorney General Eric Holder warned that risk assessment scores might be injecting bias into the courts and called for the U.S. Sentencing Commission to study their use", which is not in this line. The mapping has drifted from the narration. | block 5: this line states "twenty fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 5: this block traces to no fact at all. Its role is "turn", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (36 words), F2 (29 words), F3 (33 words), F4 (24 words), F5 (30 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  54. 11:34:38errorstage script failed: no block survived checking — block 3: POL-5 — 32 words (need 23-27). Estimated ~12.3s. | block 4: POL-5 — 38 words (need 23-27). Estimated ~14.6s. | block 5: POL-5 — 38 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~14.6s. | block 6: POL-5 — 31 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~11.9s. | block 3: this line states "two" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 5: this line states "two thousand fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (36 words), F2 (29 words), F3 (33 words), F4 (24 words), F5 (30 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  55. 11:34:38decision.scriptreject — block 3: POL-5 — 32 words (need 23-27). Estimated ~12.3s. | block 4: POL-5 — 38 words (need 23-27). Estimated ~14.6s. | block 5: POL-5 — 38 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~14.6s. | block 6: POL-5 — 31 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~11.9s. | block 3: this line states "two" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 5: this line states "two thousand fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (36 words), F2 (29 words), F3 (33 words), F4 (24 words), F5 (30 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  56. 11:22:47errorstage script failed: no block survived checking — block 3: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 5: POL-5 — 39 words (need 23-27). Estimated ~15.0s. | block 6: POL-5 — 32 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~12.3s. | block 3: the mapping quotes "Brisha Borden stole a child's bicycle and scooter valued at $80, but she has not been charged with any new crimes in the two years following her arrest", which is not in this line. The mapping has drifted from the narration. | block 3: this line states "two" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 5: the mapping quotes "The algorithm was nearly twice as likely to falsely flag black defendants as future criminals compared to white defendants", which is not in this line. The mapping has drifted from the narration. | block 6: the mapping quotes "In twenty-fourteen, U.S. Attorney General Eric Holder warned that risk assessment scores might be injecting bias into courts and called for the U.S. Sentencing Commission to study their use", which is not in this line. The mapping has drifted from the narration. | block 6: this line states "twenty-fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (36 words), F2 (33 words), F3 (33 words), F5 (29 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.
  57. 11:22:47decision.scriptreject — block 3: POL-5 — 29 words (need 23-27). Estimated ~11.2s. | block 5: POL-5 — 39 words (need 23-27). Estimated ~15.0s. | block 6: POL-5 — 32 words (need 23-27); 4 sentences (need 2-3) — sentence-end pauses are what fill the ten-second window, so one long sentence runs short. Estimated ~12.3s. | block 3: the mapping quotes "Brisha Borden stole a child's bicycle and scooter valued at $80, but she has not been charged with any new crimes in the two years following her arrest", which is not in this line. The mapping has drifted from the narration. | block 3: this line states "two" without tracing to a fact. Map it or cut it — a number on screen is a claim. | block 3: this block traces to no fact at all. Its role is "evidence", which is reporting — only the kicker may be pure framing. Map at least one assertion to a fact, or replace the block. | block 5: the mapping quotes "The algorithm was nearly twice as likely to falsely flag black defendants as future criminals compared to white defendants", which is not in this line. The mapping has drifted from the narration. | block 6: the mapping quotes "In twenty-fourteen, U.S. Attorney General Eric Holder warned that risk assessment scores might be injecting bias into courts and called for the U.S. Sentencing Commission to study their use", which is not in this line. The mapping has drifted from the narration. | block 6: this line states "twenty-fourteen" without tracing to a fact. Map it or cut it — a number on screen is a claim. (after 8 attempts — surfacing rather than retrying further) What to do: retry — refusals are free and the model redrafts. If it keeps refusing, trim F1 (36 words), F2 (33 words), F3 (33 words), F5 (29 words) to one spoken sentence — a narrated line holds 23-27 words TOTAL, and a fact that long leaves no room to say it.